ASCEND: spontaneous Chinese-English code-switching speech corpus
- Provider
- Holy Lovenia, Samuel Cahyawijaya, Genta Indra Winata et al. (CAiRE, HKUST)
- License
- CC-BY-SA-4.0 (source)
- Size
- 7 files
- Format
- Parquet (in sampled files)
- Stored at
- Hugging Face (https://huggingface.co/datasets/CAiRE/ASCEND)
- Access
- Free and public. No account with the data provider is needed.
- Category
- Speech & Audio
- Source
- https://arxiv.org/abs/2112.06223
About ASCEND: spontaneous Chinese-English code-switching speech corpus
ASCEND (A Spontaneous Chinese-English Dataset) is a high-quality corpus of spontaneous, multi-turn conversational dialogue with Chinese-English code-switching, collected in Hong Kong by the Centre for Artificial Intelligence Research (CAiRE) at HKUST. It consists of 10.62 hours of spontaneous speech in about 12.3K utterances, split 8:1:1 into training, validation and test sets with a balanced gender proportion in each. Each utterance is labelled with its transcription, language (Chinese, English or mixed), speaker id, session id and conversation topic. It is a benchmark for code-switching automatic speech recognition, language identification within utterances, and bilingual Mandarin-English speech modelling. This repository holds the corpus as five Parquet files under main/ (train, validation and test shards) with audio and the fields above (measured 2026-11-14; the Hub reports about 19 GB). Load with datasets.load_dataset('CAiRE/ASCEND'). Released by CAiRE under CC BY-SA 4.0; cite Lovenia et al., 'ASCEND: A Spontaneous Chinese-English Dataset for Code-switching in Multi-turn Conversation'.
How to load ASCEND: spontaneous Chinese-English code-switching speech corpus in Python
from datasets import load_dataset
ds = load_dataset("CAiRE/ASCEND")
print(ds)Data structure
File types in sampled files: Parquet
Files
Example file: main/test-00000-of-00001.parquet
Analyze this dataset without downloading it
On Scigantic, ASCEND: spontaneous Chinese-English code-switching speech corpus opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.
This page is also available as Markdown and as JSON.
Related Speech & Audio datasets
- Vibravox: French speech recorded with body-conduction sensors
- NSynth: Neural Audio Synthesis musical notes dataset
- Norwegian Parliamentary Speech Corpus (NPSC)
- Multilingual Spoken Words Corpus (MSWC)
- The People's Speech: 30,000-hour English speech recognition corpus
- speechocean762: non-native English pronunciation scoring corpus
- Upper Waiakea Forest Reserve Passive Acoustic Recordings of Hawaiian Forest Birds
- 2 M Belebele